NVDA估值不高且有上行空间；主要风险是垂直整合面临的监管挑战。

Nvidia confirmed Thursday morning it's acquiring HuggyFace, the New York-based platform that's become the default home for open-source AI development, and they're purchasing it for roughly $13 billion.

That's Nvidia's second largest acquisition ever, behind only the 20 billion purchase of chipmaker Gro's assets in December of 2025. And it dwarfs the seven billion dollar Melanax deal that first got Nvidia into networking back way back in 2019.

So the translation is this is Nvidia positioning itself as a champion of open models against two fronts at once. Closed proprietary labs like OpenAI and Anthropic and increasingly competitive openweight Chinese models that are cheaper to run and in some benchmarks closing that performance gap.

Now, both the CEO of HuggingFace and the CEO of Nvidia then made a coordinated argument that open models give defenders an asymmetric advantage over attackers since there's more people protecting the ecosystem than attacking it.

And I think HuggingFace, you know, kind of now bolstering up next to Nvidia when it comes to this openweight open model infrastructure. This is the future. and Nvidia just planted its foot right into it.

Yeah, and this isn't an isolated move by any stretch. Nvidia has been buying its way up the AI stack all year. The pattern here, first Nvidia owned the chips. Then it built CUDA, the software layer that locks developers into its hardware. Now it owns the marketplace where virtually every major open- source model, including Meta's Lama and Mistl's models, all live.

Now, the last point, and this is why regulators are already kind of circling around them, Nvidia's share of the AI training chip market still sits above 80%. add the dominant model distribution platform that they're now building on top of that hardware monopoly, right? I think what was it, Robert? 675 billion of of hardware monopoly revenue next year, right?

Like, this is exactly the kind of vertical stacking the EU antitrust regulators have been sniffing and trying to, you know, jump on all year long. Jensen Wong isn't shy about pushing back, though. He told G20 officials earlier this week, and I quote, "Don't regulate hypothetical, theoretical harm. Regulate actual and pragmatic harm." I know a lot of people like myself own stock in Nvidia.

Yeah, this is Nvidia moving from selling picks and shovels to owning the entire mine again. I pulled up Nvidia on Wall Street favorites.com just this morning and 97 analysts cover the stock and the median price target sits at $322.50 against a current price of around $224. That's still 43.7% upside priced in by the street and none of that reflects any of the deal we're talking about today.

Austin and you and I have both been very outspoken that we believe Nvidia is quite inexpensive at this current price. So the real risk isn't whether this deal works, it's regulatory. If Brussels or Washington decides Nvidia is using hardware dominance to lock up software distribution, too, this is the acquisition that draws that formal challenge.

So the bigger takeaway for three years, the AI trade was about who built the best chips. Nvidia just told you that the next fight is about who owns the ecosystem, the models, the developer community, the distribution rails, everyone else has to build on top of. So hardware becomes a commodity eventually platforms don't. And Nvidia just spent $13 billion making sure it never has to find out which one it is.

No, I like that uh breakdown, Robert. I think Nvidia is just Jensen Wong is just an absolute unbelievable founder. I mean, he's been the founder and CEO here of Nvidia since the late 90s. I mean, this guy has grown it into this 56 trillion business over the last 20 plus years.

And I'm glad that Nvidia is on the side of open source, open weight, own your data, own your model. Like you are the one that's responsible for this. You shouldn't have to give all your data away to use AI. And I think that that's the right side of history to be on.

Anthropic citing a $35 billion deal with Nvidia backed Lambda AI for a Hut8 developed data center in the state of Texas. except Nvidia itself is holding the lease, not Lambda AI.

Second mega deal this month for Anthropic, which signed a $45 billion deal for another Nvidia backed Neocloud called Nscale.

Nvidia is over here bankrolling all of this on both sides. And we see this sort of land grab go on where they're doing whatever they can just to keep the chips moving and the compute deals rolling.